Duct leakage testing and sealing field guide for HVAC crews

What duct leakage costs, and why anyone tests for it

Duct leakage is conditioned air escaping through the joints, the seams, and the penetrations before it ever reaches a register. The supply leak dumps the air you paid to heat or cool into a ceiling cavity. The return leak pulls dirty, unconditioned attic or plenum air into the system. Either way the room comes up short and the blower runs harder to make up the difference.

The cost shows up in three places at once. The energy: the fan burns watts moving air that never does any work. The capacity: the far rooms never reach setpoint because the air leaked out upstream of them. And the static: a leaky duct shifts where the system fights, so the static-pressure reading lies about what the registers are actually getting. A leak does not announce itself the way a bulging duct does. It just quietly steals a slice of every cubic foot the blower moves.

That is why the energy code now makes the test mandatory on the systems where leakage hurts most. Sealing is a build-side decision covered in the sheet metal fabrication guide, where the seal class is set from the pressure class and the joints are built to hold. This guide is the proof: how you measure the leakage that the build either prevented or did not, and how you find and seal what failed before the duct disappears above a ceiling.

Leakage and static pressure are two different readings

Leakage and static pressure both come from the same duct, but they answer different questions, and crews mix them up constantly. Static pressure is the resistance the running blower fights, read with a manometer while the system operates. Leakage is the air the duct loses through its envelope, measured by pressurizing a capped section to a known test pressure and reading how much air it takes to hold it.

The two are related but they are not the same number. A duct can pass a leakage test and still read high static, because the static problem is undersized duct or a bad fitting, not an open joint. The external static pressure guide chases that case from the operating side. A duct can also read normal static and leak badly, because the leak bleeds off pressure in a way a single static reading does not flag.

Read them together and they triangulate the system. Leakage tells you whether the air stays inside the duct. Static tells you whether the blower can push it through. When the airflow at the register comes back low, you want to know which one is the cause before you start cutting metal or upsizing a blower, because the fix for a leak is sealant and the fix for high static is bigger duct.

What is a SMACNA leakage class?

A SMACNA leakage class, written CL, is the duct's allowable leakage in cfm per 100 square feet of duct surface area at a 1 in. w.g. test pressure. It is a tightness rating: a lower CL is a tighter duct, because it leaks fewer cfm per 100 square feet at the same pressure. CL 4 leaks less than CL 12, which leaks less than CL 24.

The class describes the construction, not a single joint. SMACNA assigns leakage classes to the way the duct is built and sealed, so a well-sealed rectangular metal duct historically targeted a tighter class than a loosely sealed one, and round and oval duct reach a lower class than rectangular for the same effort because there is less linear seam to leak. The fabrication guide ties the class back to the seam, the joint, and the seal that produce it.

Two duct envelopes change which class is realistic. Rectangular metal duct has long longitudinal seams and many transverse joints, so it leaks more for the same care. Round and spiral duct seal at couplings and have a helical lock seam that is both seam and stiffener, so it lands tighter. When the spec names a leakage class, it is naming the tightness the construction has to deliver and the test has to confirm.

How do you calculate allowable duct leakage?

The SMACNA allowable leakage uses the leakage class, the test pressure, and the duct surface area. The leakage factor F, in cfm per 100 square feet, is the leakage class times the test pressure raised to the 0.65 power: F equals CL times P to the 0.65. Multiply that factor by the duct surface area in hundreds of square feet and you have the total allowable leakage in cfm for the section under test.

The 0.65 exponent is the part people drop. Leakage does not climb in a straight line with pressure, so you cannot just scale the class by the pressure ratio. At a 1 in. test the factor equals the class exactly, because 1 to any power is 1. At higher test pressures the factor grows on the 0.65 curve, so a duct tested at 4 in. leaks more per 100 square feet than the same duct tested at 1 in., and the allowable has to grow with it.

SMACNA bases the allowable on surface area rather than a flat percentage of fan airflow, and that choice matters on a big system. A percentage of fan flow lets a large duct hide a large absolute leak. The surface-area basis makes the allowable scale with how much duct you actually built, so a sprawling system carries a proportionally larger allowance and a small one carries a smaller one. Verify the leakage class, the test pressure, and the area basis against the project spec and the current SMACNA edition before you sign off on a number.

CL

Leakage class, the allowable leakage in cfm per 100 sq ft at 1 in. w.g.

P

Test static pressure in inches of water gauge, commonly the design pressure class

A

Duct surface area of the section under test, in square feet

Field example: working an allowable and a pass/fail

Take a high-pressure rectangular supply run built to a 4 in. w.g. pressure class, with 2,500 square feet of duct surface in the section under test, held to leakage class CL 4 by the spec because the energy code put it in scope. Test it at the 4 in. design pressure.

Run the numbers. The pressure factor is 4 to the 0.65 power, about 2.46, so the leakage factor F is 4 times 2.46, which is 9.8 cfm per 100 square feet. The section is 2,500 square feet, or 25 hundreds of square feet, so the allowable is 9.8 times 25, about 246 cfm. If the system moves 12,000 cfm, that allowable is roughly 2 percent of the fan airflow, which is the sanity check that keeps the surface-area math honest.

Now read the test. The calibrated fan needs 180 cfm to hold the section at 4 in., so the duct leaks 180 cfm against a 246 cfm allowance. That passes, with margin. Push the same duct and the reading comes back at 310 cfm, and it fails by a clear 64 cfm, which is real, findable leakage, not measurement noise. The number tells you which way to walk: 180 means cap it and move on, 310 means get the smoke and find the open joints.

How do you test ductwork for leakage?

You isolate a section, pressurize it with a calibrated fan to the test pressure, and measure the cfm it takes to hold that pressure. That make-up airflow is the leakage, because in a sealed, steady section every cubic foot the fan adds is a cubic foot escaping through the envelope. The SMACNA HVAC Air Duct Leakage Test Manual is the method behind the work.

The setup is specific. Cap every opening in the section: branch stubs, takeoffs, the open ends, and any device opening, using rigid caps or sealed plugs, not loose plastic that balloons and reads as leakage itself. Connect the test fan through a calibrated orifice plate or flow meter, because the orifice is what turns a pressure drop across a known hole into an accurate cfm reading. Run a tube from the duct to a manometer to read the static inside the section.

Then bring it up to pressure. Start the fan, throttle it until the section sits at the test pressure called out by the spec, commonly the design pressure class, and let it stabilize before you read. Record the leakage cfm off the orifice at that held pressure. The test runs the same way on the positive side and the negative side, since a duct can leak under suction as well as under pressure, and the return side often runs negative. Compare the measured cfm to the allowable from the leakage-class calculation, and that comparison is the whole result.

What ductwork has to be leak tested?

Under ASHRAE 90.1, the trigger is pressure and exposure. Duct designed to operate above 3 in. w.g., and all duct located outdoors regardless of pressure, has to be leak tested. Duct at 3 in. w.g. or below indoors generally does not carry a mandatory test under 90.1, though a tighter project spec can always require one the code does not. The IECC carries parallel language. Confirm the trigger against the adopted code edition and any local amendments.

You usually test a sample, not every foot. 90.1 lets the building owner or the owner's representative pick representative sections totaling at least 25 percent of the installed duct by surface area, so the sample has to be a real cross-section of the work, not the easiest run to cap. A smart owner picks sections that include the joints and the seams the trade is most likely to have rushed, because a sample stacked with the good work proves nothing.

The duct most worth testing is the high-pressure supply, the outdoor-air and exhaust duct exposed to weather, and any run feeding a space that cannot tolerate a shortfall. On these, a few percent of leakage is a large absolute volume of conditioned air leaving the system every hour the fan runs. The energy code put the threshold at 3 in. for exactly this reason: that is where leakage stops being a rounding error and starts being a measurable loss.

Sealing to the seal class

Sealing is closing the leak paths the air finds, and SMACNA grades how much you close by seal class. Seal Class A seals all transverse joints, all longitudinal seams, and all duct wall penetrations, the openings made by pipes, conduit, tie rods, and wires. Seal Class B seals the transverse joints and the longitudinal seams. Seal Class C seals the transverse joints only. The historic pairing put Class C around 2 in. w.g., Class B around 3 in., and Class A at 4 in. and up.

Current practice has moved tighter. ASHRAE 90.1 calls for duct and plenums with a pressure-class rating to be built to Seal Class A, so on energy-code work the answer is usually seal everything, not the lighter classes the older pressure pairing allowed. The seal class is set on the build side and detailed in the fabrication guide; confirm the required class against the project spec and the current SMACNA edition before the crew picks up a mastic brush.

The blunt rule that makes or breaks the leakage test: seal before it is hidden, not after. A joint sealed as the duct goes up is a joint you can reach and tool properly. A joint you try to seal after the next section is hung, the insulation is on, and the ceiling is closing is a joint you reach halfway and seal worse. Crews that defer sealing to a cleanup pass are the crews whose sample fails the test, because the leaks are exactly the joints that became unreachable.

Mastic, tape, and gaskets: what actually seals

Duct mastic is the durable seal. Brushed or troweled over the joint at the thickness the manufacturer calls for, with fabric mesh tape embedded where the gap is wider than mastic alone will bridge, it cures to a flexible seal that moves with the duct and holds for the life of the system. It is messy, it is slow next to slapping on a roll of tape, and it is what passes the test.

Pressure-sensitive tape has a place, but a narrow one. Tape used as a primary sealant has to be listed to UL 181A for rigid duct or UL 181B for flexible duct connectors, applied to a clean, dry surface, and used inside its listing. Foil tape with the proper listing, rubbed down hard, can seal. The catch is that any tape lets go as it ages if the surface was dusty or oily when it went on, and then the joint reopens behind the insulation where nobody looks.

Gaskets seal the flanged joints. A continuous gasket laid into a TDC, TDF, or bar-flange connection seals metal to metal where bolts and cleats alone would leak, and it has to be unbroken at the corners, which is exactly where a hurried crew leaves a gap. And the thing every pro already knows: cloth-backed duct tape, the gray hardware-store roll, is not a duct sealant. It dries out, the adhesive fails, and it peels. Despite the name, it is the one product that does not belong on a duct joint you expect to pass a test.

Finding the leak when a section fails

A failed test gives you a number, not a location, so the next job is finding where the air leaves. The fastest method is to leave the test fan running so the section stays at pressure, then work the duct by hand. A leak under positive pressure pushes a stream of air you can feel by running a wet hand or the back of your fingers along every joint, seam, and penetration. The corners of transverse joints and the ends of longitudinal seams are where it usually is.

When the hand is not enough, add smoke. A smoke pencil or a theatrical-smoke machine fed into the pressurized duct streams out of every leak path visibly, and on a large section that is the difference between an afternoon of guessing and twenty minutes of finding. On a negative-pressure test the smoke gets pulled in at the leaks instead of pushed out, so you watch for the draw rather than the plume.

Listen, too. A real leak at pressure hisses, and on a quiet site you can walk a trunk and hear the open joints before you feel them. Mark every leak as you find it rather than trying to remember them, because a section that failed by 60 cfm usually has several small leaks adding up, not one obvious hole, and the one you skip is the one that fails the re-test.

Failing, re-sealing, and re-testing

A failed leakage test is not a verdict on the duct, it is a punch list. The section leaked more than the allowable, so you find the leaks, seal them, and test again. The order matters: keep the section capped and at pressure while you hunt, mark every leak, then drop the pressure, seal them all, give the mastic the cure time the product calls for, and re-pressurize.

Resist the temptation to seal one obvious leak and re-test on hope. The failure is almost always several leaks summed, so a partial seal gives you a second failure and a second mobilization. Walk the whole section, fix everything you marked, and only then bring the fan back. Document both runs, the failing number and the passing number, because the record of the re-test is what proves the section was actually brought into spec rather than waved through.

The expensive version of this is the test that happens after the ceiling closed. Now finding and sealing a leak means demolition, and a section that should have cost an hour of mastic costs a day of teardown. Test while every joint is still reachable, fail cheap, and re-seal before the duct disappears.

Tighter construction leaks less to begin with

The cheapest leakage to seal is the leakage the build never created. A duct assembled with the right seam for its pressure, gasketed flanged joints where the class calls for them, and full drive engagement on the slip joints starts the test most of the way to a pass before a brush of mastic touches it. The fabrication guide ties each of those choices to the pressure class.

The opposite is a duct that fights you. A snap-lock seam pushed past its pressure range works open along its whole length, a slip-and-drive with the drives half-engaged leaks at every corner, and a flanged joint run metal-to-metal without its gasket leaks no matter how hard the bolts are torqued. Sealant over a bad joint is a patch on a problem the construction should have prevented, and patched joints are the ones that reopen.

So the move on a tight-leakage job is to spec the construction up front rather than counting on sealant to rescue it. Gasketed four-bolt flanges on the high-pressure runs, round and oval where the architecture allows, and seams matched to the class. A duct built tight tests tight, and a duct built loose leaks faster than a crew can seal it.

High-pressure, data center, and critical air systems

On high-pressure central systems, data center cooling, and critical air handling, leakage stops being an efficiency line item and becomes an airflow the facility cannot lose. These systems run the higher pressure classes, where the energy code puts them squarely in the mandatory-test bracket, and they carry airflow the building has budgeted to the cubic foot. A few percent of leakage on a large high-pressure system is a large absolute volume leaving the duct every hour, and a measurable load on the fan.

Two things tighten on these jobs. The leakage class gets stricter and is verified by test rather than assumed, because the absolute loss is what the design cannot tolerate, not the percentage. And the joints lean to the flanged, gasketed connections that seal repeatably at pressure, because a slip joint that is fine at 1 in. is the wrong call at 6 in. The project spec on critical work is usually tighter than the SMACNA default and is the document that controls.

Data center supply is the case where this bites hardest, because cooling air leaked out of a high-pressure run upstream of the racks is air the servers never see, and the only symptom is a hot aisle nobody can explain from the unit. Test it, seal it, and document it, because on these systems an untested duct is an airflow risk the commissioning agent will not sign off.

What sealing the duct buys back

A tight duct buys back fan energy and capacity, and on a metered system you can put numbers to it. Air that stays in the duct is air the fan does not have to over-supply to make up a leak, so the blower holds its design airflow at a lower pressure and draws fewer watts every hour it runs. Over a cooling season that is real money on the bill, which is exactly why the energy code wrote the test into 90.1 and the IECC.

Capacity is the other half. A supply leak in an unconditioned plenum throws away cooling or heating that never reaches the room, so the far spaces underperform and somebody turns the equipment up or oversizes the next one. Seal the leak and the same equipment serves more space, because the air it makes actually arrives. On a return leak the win is air quality too, since a sealed return stops pulling dusty, humid, unconditioned air into the system.

The honest framing is that sealing rarely pays for itself in comfort alone on a small house, where the leak is small and the runs are short. It pays clearly on the bigger, higher-pressure, longer-run systems, which is the same set of systems the code already makes you test. The places the test is mandatory are the places the savings are real.

The commissioning report and the pass/fail record

The leakage test only counts if it is written down. The commissioning report turns a fan reading into a defensible record: the section tested, its surface area, the test pressure, the leakage class and the allowable cfm it produces, the measured leakage, and the plain pass or fail. The commissioning agent and the inspector both read this to confirm the duct met the spec before it was buried.

Tie the report to the sample. Because you test a representative fraction, the report has to identify which sections were tested and show they really cover at least the required percentage of the installed surface area, so the record stands up to the question of whether the sample was honest. A report that quietly tested only the easy runs is a report that proves nothing about the duct that failed.

Record the failures and the re-tests, not just the final pass. The value of the document six months out is that it shows the duct was brought into spec, including the sections that needed re-sealing, so when an airflow complaint comes in later the report answers whether the duct was ever tight. A pass with no history behind it is a bare number, with nothing to show how the duct got there.

What the owner inherits

A duct that passed at commissioning does not stay sealed by itself, and the owner inherits the slow ways it loosens. Tape that was used where mastic should have been lets go over a few seasons. A duct that sags between hangers opens its joints. A flex tail that gets crushed during another trade's work throttles a branch. The leakage that the test caught and sealed can creep back if the system is abused.

The handover that helps the owner is the test record plus the access. Leave the commissioning report so the next contractor knows what the duct was supposed to leak, and make sure the access doors that the build needed for dampers and devices are sealed and labeled, because an unsealed access door is a leak the owner pays for every day. A duct re-tested years later against its original report tells the owner whether the system drifted or held.

The maintenance lesson is short. Seal it right the first time with mastic and listed materials, and the duct stays tight with no upkeep. Seal it with the wrong tape to make a deadline and the owner inherits a leak that reappears the year after everyone has left the job.

What to document

Skip the paperwork and you run the leakage test over, because an unlogged result and an unrecorded seal carry no weight when someone later asks how tight this duct really is. The record is what answers the question later: was this duct ever tight, and to what class. Capture it per section so the next reader can reproduce the result instead of taking it on faith.

For each tested section, record the duct type and pressure class, the surface area, the seal class and the sealing method used, the leakage class targeted, the test pressure, the allowable leakage the calc produced, the measured leakage, the pass or fail, any re-seal and re-test, and the instrument and date. The table below is the minimum that makes a result defensible to a commissioning agent or an inspector.

Common mistakes

• Skipping the leakage test where the code requires it, then chasing the lost airflow from the wrong end.
• Using cloth-backed duct tape, or any tape not listed to UL 181A or 181B, as a primary sealant.
• Sealing the joints after the duct is insulated and hung, so the leaking joints can no longer be reached.
• Sealing to the wrong seal class, leaving seams or penetrations open on duct that called for Class A.
• Letting the leaks at the takeoffs, the corners, and the longitudinal seam ends go because the body of the duct looks tight.
• Testing at the wrong pressure, which moves the allowable up or down off the 0.65 curve and gives a false pass or fail.
• Capping a section with loose plastic that balloons and reads as leakage it does not have.
• Sealing one obvious leak after a failure and re-testing on hope instead of finding all of them.
• Testing only the easy runs to make the 25 percent sample, so the report proves nothing about the rushed work.

Field checklist

Standards and references

SMACNA owns the leakage side of duct work. The HVAC Air Duct Leakage Test Manual gives the leakage classes, the leakage factor F equals CL times P to the 0.65, and the test method using a calibrated fan and orifice. The HVAC Duct Construction Standards set the pressure classes, the seam and joint construction, and the seal classes A, B, and C that produce a given leakage class. Cite these by topic and confirm the numbers against the edition the job is built to, because the tables move between editions.

The energy code makes the test mandatory. ANSI/ASHRAE/IES Standard 90.1 requires leakage testing on duct designed above 3 in. w.g. and on all outdoor duct, sets the maximum permissible leakage using a leakage class of 4 for round and rectangular, calls for Seal Class A on pressure-class duct, and lets the owner select representative sections totaling at least 25 percent of the installed surface area. The IECC carries parallel requirements. The adopted edition and local amendments control, so verify rather than quoting a remembered figure.

Closure materials are governed by their listings. Pressure-sensitive and heat-applied tapes used as duct sealant are listed to UL 181A for rigid duct and UL 181B for flexible connectors, and mastics carry their own listings and application limits. Above all of these, the project specification governs, and on high-pressure and critical systems it is usually stricter than the SMACNA default and is the document that controls the leakage class and the test.

Units, terms, and conversions

Leakage work crosses a few units and a lot of shorthand, so the same idea reads differently across a drawing, a spec, and a test report.

Test pressure is in inches of water gauge, written in. w.g. or in. wc, and one inch of water gauge is about 249 pascals. Leakage is in cfm, cubic feet per minute, and the leakage class CL is normalized to cfm per 100 square feet of duct surface at a 1 in. w.g. test. The leakage factor F is the class scaled to the test pressure by F equals CL times P to the 0.65. Seal class A, B, or C describes what gets sealed, while leakage class is the measured tightness; the two are related but they are not the same thing.

Leakage class (CL)

Allowable leakage in cfm per 100 sq ft at 1 in. w.g.; a lower number is a tighter duct

Leakage factor (F)

Allowable cfm per 100 sq ft at the test pressure, F = CL x P^0.65

Seal class (A/B/C)

What must be sealed: A is everything, B is joints and seams, C is transverse joints only

Test pressure (P)

The static the section is pressurized to for the test, in inches of water gauge

Orifice plate

The calibrated restriction that turns a pressure drop into an accurate leakage cfm reading

UL 181A / 181B

The listings for tapes and closures used on rigid duct (181A) and flexible connectors (181B)

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• What is a SMACNA pressure class?
• How do I measure total external static pressure?
• What is an acceptable duct leakage class?
• How do you size a gutter?
• Does duct cleaning really work?